Create trans_network object for co-occurrence network analysis.
This class is a wrapper for a series of network analysis methods, including the network construction approaches, network attributes analysis, eigengene analysis, network subsetting, node and edge properties extraction, network plotting, and other network operations.
new()
This function is used to create the trans_network object, store the important intermediate data and calculate correlations if cal_cor parameter is selected.
trans_network$new(
dataset = NULL,
cor_method = c("pearson", "spearman", "kendall")[1],
cal_cor = c("base", "WGCNA", "SparCC", NULL)[1],
taxa_level = "OTU",
filter_thres = 0,
nThreads = 1,
SparCC_simu_num = 100,
env_cols = NULL,
add_data = NULL
)datasetthe object of microtable Class.
cor_methoddefault "pearson"; "pearson", "spearman" or "kendall"; correlation algorithm, only use for correlation-based network.
cal_cordefault "base"; "base", "WGCNA", "SparCC" or NULL; correlation method; NULL denote no correlation calculation, used for non-correlation based network, such as SpiecEasi and FlashWeave methods.
taxa_leveldefault "OTU"; taxonomic rank; 'OTU' denotes using feature table directly; other available options should be one of the colnames of microtable$tax_table.
filter_thresdefault 0; the relative abundance threshold.
nThreadsdefault 1; the thread number used for "WGCNA" and SparCC.
SparCC_simu_numdefault 100; SparCC simulation number for bootstrap.
env_colsdefault NULL; numeric or character vector to select the column names of environmental data in dataset$sample_table; the environmental data can be used in the correlation network (as the nodes) or FlashWeave network.
add_datadefault NULL; provide environmental table additionally instead of env_cols parameter; rownames must be sample names.
res_cor_p list; include the correlation matrix and p value matrix.
\donttest{
data(dataset)
# for correlation network
t1 <- trans_network$new(dataset = dataset, cal_cor = "base",
taxa_level = "OTU", filter_thres = 0.0001)
# for other network
t1 <- trans_network$new(dataset = dataset, cal_cor = NULL)
}
cal_network()
Calculate network based on the correlation method or SpiecEasi package or julia FlashWeave package or beemStatic package.
trans_network$cal_network(
network_method = c("COR", "SpiecEasi", "FlashWeave", "beemStatic")[1],
COR_p_thres = 0.01,
COR_p_adjust = "fdr",
COR_weight = TRUE,
COR_cut = 0.6,
COR_optimization = FALSE,
COR_optimization_low_high = c(0.4, 0.8),
SpiecEasi_method = "mb",
FlashWeave_tempdir = NULL,
FlashWeave_meta_data = FALSE,
FlashWeave_other_para = "alpha=0.01,sensitive=true,heterogeneous=true",
beemStatic_t_strength = 0.001,
beemStatic_t_stab = 0.8,
add_taxa_name = "Phylum",
usename_rawtaxa_when_taxalevel_notOTU = FALSE,
...
)network_methoddefault "COR"; "COR", "SpiecEasi", "FlashWeave" or "beemStatic"; The option details:
correlation-based network; use the correlation and p value matrixes in object$res_cor_p returned from trans_network$new; See Deng et al. (2012) <doi:10.1186/1471-2105-13-113> for other details
SpiecEasi network; relies on algorithms for sparse neighborhood and inverse covariance selection; belong to the category of conditional dependence and graphical models; see https://github.com/zdk123/SpiecEasi for installing the R package; see Kurtz et al. (2015) <doi:10.1371/journal.pcbi.1004226> for the algorithm details
FlashWeave network; Local-to-global learning framework; belong to the category of conditional dependence and graphical models; good performance on heterogenous datasets to find direct associations among taxa; see https://github.com/meringlab/FlashWeave.jl for installing julia language and FlashWeave package; julia must be in the computer system env path, otherwise the program can not find julia; see Tackmann et al. (2019) <doi:10.1016/j.cels.2019.08.002> for the algorithm details
beemStatic network; extend generalized Lotka-Volterra model to cases of cross-sectional datasets to infer interaction among taxa based on expectation-maximization algorithm; see https://github.com/CSB5/BEEM-static for installing the R package; see Li et al. (2021) <doi:10.1371/journal.pcbi.1009343> for algorithm details
COR_p_thresdefault 0.01; the p value threshold for the correlation-based network.
COR_p_adjustdefault "fdr"; p value adjustment method, see method of p.adjust function for available options.
COR_weightdefault TRUE; whether use correlation coefficient as the weight of edges; FALSE represents weight = 1 for all edges.
COR_cutdefault 0.6; correlation coefficient threshold for the correlation network.
COR_optimizationdefault FALSE; whether use random matrix theory (RMT) based method to determine the correlation coefficient; see https://doi.org/10.1186/1471-2105-13-113
COR_optimization_low_highdefault c(0.4, 0.8); the low and high value threshold used for the RMT optimization; only useful when COR_optimization = TRUE.
SpiecEasi_methoddefault "mb"; either 'glasso' or 'mb';see spiec.easi function in package SpiecEasi and https://github.com/zdk123/SpiecEasi.
FlashWeave_tempdirdefault NULL; The temporary directory used to save the temporary files for running FlashWeave; If not assigned, use the system user temp.
FlashWeave_meta_datadefault FALSE; whether use env data for the optimization, If TRUE, the function automatically find the object$env_data in the object and generate a file for meta_data_path parameter of FlashWeave.
FlashWeave_other_paradefault "alpha=0.01,sensitive=true,heterogeneous=true"; the parameters used for FlashWeave; user can change the parameters or add more according to FlashWeave help document; An exception is meta_data_path parameter as it is generated based on the data inside the object, see FlashWeave_meta_data parameter for the description.
beemStatic_t_strengthdefault 0.001; for network_method = "beemStatic"; the threshold used to limit the number of interactions (strength); same with the t.strength parameter in showInteraction function of beemStatic package.
beemStatic_t_stabdefault 0.8; for network_method = "beemStatic"; the threshold used to limit the number of interactions (stability); same with the t.stab parameter in showInteraction function of beemStatic package.
add_taxa_namedefault "Phylum"; one or more taxonomic rank name; used to add taxonomic rank name to network node properties.
usename_rawtaxa_when_taxalevel_notOTUdefault FALSE; whether replace the name of nodes using the taxonomic information.
...paremeters pass to spiec.easi function of SpiecEasi package when network_method = "SpiecEasi" or func.EM function of beemStatic package when network_method = "beemStatic".
res_network stored in object.
\donttest{
# for correlation network
t1 <- trans_network$new(dataset = dataset, cal_cor = "base",
taxa_level = "OTU", filter_thres = 0.001)
t1$cal_network(COR_p_thres = 0.05, COR_cut = 0.6)
t1 <- trans_network$new(dataset = dataset, cal_cor = NULL, filter_thres = 0.003)
t1$cal_network(network_method = "SpiecEasi", SpiecEasi_method = "mb")
t1 <- trans_network$new(dataset = dataset, cal_cor = NULL, taxa_level = "OTU", filter_thres = 0.005)
t1$cal_network(network_method = "beemStatic")
t1 <- trans_network$new(dataset = dataset, cal_cor = NULL, filter_thres = 0.001)
t1$cal_network(network_method = "FlashWeave")
}
cal_module()
Calculate network modules and add module names to the network node properties.
trans_network$cal_module( method = "cluster_fast_greedy", module_name_prefix = "M" )
methoddefault "cluster_fast_greedy"; the method used to find the optimal community structure of a graph; the following are available functions (options) from igraph package: "cluster_fast_greedy", "cluster_optimal", "cluster_edge_betweenness", "cluster_infomap", "cluster_label_prop", "cluster_leading_eigen", "cluster_louvain", "cluster_spinglass", "cluster_walktrap". For the details of these functions, see the help document, such as help(cluster_fast_greedy); Note that the default "cluster_fast_greedy" method can only be used for undirected network. If the user selects network_method = "beemStatic" in cal_network function or provides other directed network, please use cluster_optimal or others for the modules identification.
module_name_prefixdefault "M"; the prefix of module names; module names are made of the module_name_prefix and numbers; numbers are assigned according to the sorting result of node numbers in modules with decreasing trend.
res_network with modules, stored in object.
\donttest{
t1 <- trans_network$new(dataset = dataset, cal_cor = "base",
taxa_level = "OTU", filter_thres = 0.0002)
t1$cal_network(COR_p_thres = 0.01, COR_cut = 0.6)
t1$cal_module(method = "cluster_fast_greedy")
}
save_network()
Save network as gexf style, which can be opened by Gephi (https://gephi.org/).
trans_network$save_network(filepath = "network.gexf")
filepathdefault "network.gexf"; file path to save the network.
None.
\dontrun{
t1$save_network(filepath = "network.gexf")
}
cal_network_attr()
Calculate network properties.
trans_network$cal_network_attr()
res_network_attr stored in object.
\donttest{
t1$cal_network_attr()
}
cal_node_type()
Calculate node properties. This function will be deprecated in the next release! Please use get_node_table function!
trans_network$cal_node_type()
see the Return part in function get_node_table.
get_node_table()
Get the node property table. The properties may include the node names, modules allocation, degree, betweenness, abundance, taxonomy, within-module connectivity and among-module connectivity <doi:10.1016/j.geoderma.2022.115866>.
Authors: Chi Liu, Umer Zeeshan Ijaz
trans_network$get_node_table(node_roles = TRUE)
node_rolesdefault TRUE; whether calculate node roles, i.e. Module hubs, Network hubs, Connectors and Peripherals <doi:10.1016/j.geoderma.2022.115866>.
res_node_table in object; Abundance expressed as a percentage; z denotes within-module connectivity; p denotes among-module connectivity.
\donttest{
t1$get_node_table(node_roles = TRUE)
}
get_edge_table()
Get the edge property table, including connected nodes, label and weight.
trans_network$get_edge_table()
res_edge_table in object.
\donttest{
t1$get_edge_table()
}
get_adjacency_matrix()
Get the adjacency matrix from the network graph.
trans_network$get_adjacency_matrix(...)
...parameters passed to as_adjacency_matrix function of igraph package.
res_adjacency_matrix in object.
\donttest{
t1$get_adjacency_matrix(attr = "weight")
}
plot_network()
Plot the network based on a series of methods from other packages, such as igraph, ggraph and networkD3. The networkD3 package provides dynamic network. It is especially useful for a glimpse of the whole network structure and finding the interested nodes and edges in a large network. In contrast, the igraph and ggraph methods are suitable for relatively small network.
trans_network$plot_network(
method = c("igraph", "ggraph", "networkD3")[1],
node_label = "name",
node_color = NULL,
ggraph_layout = "fr",
ggraph_node_size = 2,
ggraph_text_color = NULL,
ggraph_text_size = 3,
networkD3_node_legend = TRUE,
networkD3_zoom = TRUE,
...
)methoddefault "igraph"; The available options:
call plot.igraph function in igraph package for a static network; see plot.igraph for the parameters
call ggraph function in ggraph package for a static network
use forceNetwork function in networkD3 package for a dynamic network; see forceNetwork function for the parameters
node_labeldefault "name"; node label shown in the plot for method = "ggraph" or method = "networkD3"; Please see the column names of object$res_node_table, which is the returned table of function object$get_node_table; User can select other column names in res_node_table.
node_colordefault NULL; node color assignment for method = "ggraph" or method = "networkD3"; Select a column name of object$res_node_table, such as "module".
ggraph_layoutdefault "fr"; for method = "ggraph"; see layout parameter of create_layout function in ggraph package.
ggraph_node_sizedefault 2; for method = "ggraph"; the node size.
ggraph_text_colordefault NULL; for method = "ggraph"; a column name of object$res_node_table; User can select other column names or change the content of object$res_node_table.
ggraph_text_sizedefault 3; for method = "ggraph"; the node label text size.
networkD3_node_legenddefault TRUE; used for method = "networkD3"; logical value to enable node colour legends; Please see the legend parameter in networkD3::forceNetwork function.
networkD3_zoomdefault TRUE; used for method = "networkD3"; logical value to enable (TRUE) or disable (FALSE) zooming; Please see the zoom parameter in networkD3::forceNetwork function.
...parameters passed to plot.igraph function when method = "igraph" or forceNetwork function when method = "networkD3".
network plot.
\donttest{
t1$plot_network(method = "igraph", layout = layout_with_kk)
t1$plot_network(method = "ggraph", node_color = "module")
t1$plot_network(method = "networkD3", node_color = "module")
}
cal_eigen()
Calculate eigengenes of modules, i.e. the first principal component based on PCA analysis, and the percentage of variance <doi:10.1186/1471-2105-13-113>.
trans_network$cal_eigen()
res_eigen and res_eigen_expla in object.
\donttest{
t1$cal_eigen()
}
plot_taxa_roles()
Plot the classification and importance of nodes, see object$res_node_table for the variable names used in the parameters.
trans_network$plot_taxa_roles(
use_type = c(1, 2)[1],
roles_color_background = FALSE,
roles_color_values = NULL,
plot_module = FALSE,
x_lim = c(0, 1),
use_level = "Phylum",
show_value = c("z", "p"),
show_number = 1:10,
plot_color = "Phylum",
plot_shape = "taxa_roles",
plot_size = "Abundance",
color_values = RColorBrewer::brewer.pal(12, "Paired"),
shape_values = c(16, 17, 7, 8, 15, 18, 11, 10, 12, 13, 9, 3, 4, 0, 1, 2, 14),
...
)use_typedefault 1; 1 or 2; 1 represents taxa roles area plot; 2 represents the layered plot with taxa as x axis.
roles_color_backgrounddefault FALSE; for use_type=1; TRUE: use background colors for each area; FALSE: use classic point colors.
roles_color_valuesdefault NULL; for use_type=1; color palette for background or points.
plot_moduledefault FALSE; for use_type=1; whether plot the modules information.
x_limdefault c(0, 1); for use_type=1; x axis range when roles_color_background = FALSE.
use_leveldefault "Phylum"; for use_type=2; used taxonomic level in x axis.
show_valuedefault c("z", "p"); for use_type=2; used variable in y axis.
show_numberdefault 1:10; for use_type=2; showed number in x axis, sorting according to the nodes number.
plot_colordefault "Phylum"; for use_type=2; used variable for color.
plot_shapedefault "taxa_roles"; for use_type=2; used variable for shape.
plot_sizedefault "Abundance"; for use_type=2; used for point size; a fixed number (e.g. 5) is also available.
color_valuesdefault RColorBrewer::brewer.pal(12, "Paired"); for use_type=2; color vector
shape_valuesdefault c(16, 17, 7, 8, 15, 18, 11, 10, 12, 13, 9, 3, 4, 0, 1, 2, 14); for use_type=2; shape vector, see ggplot2 tutorial for the shape meaning.
...paremeters pass to geom_point.
ggplot.
\donttest{
t1$plot_taxa_roles(roles_color_background = FALSE)
}
subset_network()
Subset of the network.
trans_network$subset_network(node = NULL, edge = NULL, rm_single = TRUE)
nodedefault NULL; provide the node names that you want to use in the sub-network.
edgedefault NULL; provide the edge name needed; must be one of "+" or "-".
rm_singledefault TRUE; whether remove the nodes without any edge in the sub-network.
a new network
\donttest{
t1$subset_network(node = t1$res_node_table %>% base::subset(module == "M1") %>%
rownames, rm_single = TRUE)
# return a sub network that contains all nodes of module M1
}
cal_powerlaw()
Fit degrees to a power law distribution. First, perform a bootstrapping hypothesis test to determine whether degrees follow a power law distribution. If the distribution follows power law, then fit degrees to power law distribution and return the parameters.
trans_network$cal_powerlaw(...)
...paremeters pass to fit_power_law function in igraph package.
res_powerlaw_p and res_powerlaw_fit; see bootstrap_p function in poweRlaw package for the bootstrapping p value details; see fit_power_law function in igraph package for the power law fit return details.
\donttest{
t1$cal_powerlaw()
}
trans_comm()
Transform classifed features to community-like microtable object for further analysis, such as module-taxa table.
trans_network$trans_comm(use_col = "module", abundance = TRUE)
use_coldefault "module"; which column to use as the 'community'; must be one of the name of res_node_table from function get_node_table.
abundancedefault TRUE; whether sum abundance of taxa. TRUE: sum the abundance for a taxon across all samples; FALSE: sum the frequency for a taxon across all samples.
a new microtable class.
\donttest{
t2 <- t1$trans_comm(use_col = "module")
}
print()
Print the trans_network object.
trans_network$print()
clone()
The objects of this class are cloneable with this method.
trans_network$clone(deep = FALSE)
deepWhether to make a deep clone.
## ------------------------------------------------ ## Method `trans_network$new` ## ------------------------------------------------ data(dataset) # for correlation network t1 <- trans_network$new(dataset = dataset, cal_cor = "base", taxa_level = "OTU", filter_thres = 0.0001) # for other network t1 <- trans_network$new(dataset = dataset, cal_cor = NULL) ## ------------------------------------------------ ## Method `trans_network$cal_network` ## ------------------------------------------------ # for correlation network t1 <- trans_network$new(dataset = dataset, cal_cor = "base", taxa_level = "OTU", filter_thres = 0.001) t1$cal_network(COR_p_thres = 0.05, COR_cut = 0.6) t1 <- trans_network$new(dataset = dataset, cal_cor = NULL, filter_thres = 0.003) t1$cal_network(network_method = "SpiecEasi", SpiecEasi_method = "mb") t1 <- trans_network$new(dataset = dataset, cal_cor = NULL, taxa_level = "OTU", filter_thres = 0.005) t1$cal_network(network_method = "beemStatic") t1 <- trans_network$new(dataset = dataset, cal_cor = NULL, filter_thres = 0.001) t1$cal_network(network_method = "FlashWeave") ## ------------------------------------------------ ## Method `trans_network$cal_module` ## ------------------------------------------------ t1 <- trans_network$new(dataset = dataset, cal_cor = "base", taxa_level = "OTU", filter_thres = 0.0002) t1$cal_network(COR_p_thres = 0.01, COR_cut = 0.6) t1$cal_module(method = "cluster_fast_greedy") ## ------------------------------------------------ ## Method `trans_network$save_network` ## ------------------------------------------------ ## Not run: t1$save_network(filepath = "network.gexf") ## End(Not run) ## ------------------------------------------------ ## Method `trans_network$cal_network_attr` ## ------------------------------------------------ t1$cal_network_attr() ## ------------------------------------------------ ## Method `trans_network$get_node_table` ## ------------------------------------------------ t1$get_node_table(node_roles = TRUE) ## ------------------------------------------------ ## Method `trans_network$get_edge_table` ## ------------------------------------------------ t1$get_edge_table() ## ------------------------------------------------ ## Method `trans_network$get_adjacency_matrix` ## ------------------------------------------------ t1$get_adjacency_matrix(attr = "weight") ## ------------------------------------------------ ## Method `trans_network$plot_network` ## ------------------------------------------------ t1$plot_network(method = "igraph", layout = layout_with_kk) t1$plot_network(method = "ggraph", node_color = "module") t1$plot_network(method = "networkD3", node_color = "module") ## ------------------------------------------------ ## Method `trans_network$cal_eigen` ## ------------------------------------------------ t1$cal_eigen() ## ------------------------------------------------ ## Method `trans_network$plot_taxa_roles` ## ------------------------------------------------ t1$plot_taxa_roles(roles_color_background = FALSE) ## ------------------------------------------------ ## Method `trans_network$subset_network` ## ------------------------------------------------ t1$subset_network(node = t1$res_node_table %>% base::subset(module == "M1") %>% rownames, rm_single = TRUE) # return a sub network that contains all nodes of module M1 ## ------------------------------------------------ ## Method `trans_network$cal_powerlaw` ## ------------------------------------------------ t1$cal_powerlaw() ## ------------------------------------------------ ## Method `trans_network$trans_comm` ## ------------------------------------------------ t2 <- t1$trans_comm(use_col = "module")
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.